Since it has become open source, there seems to considerable momentum building around the OpenPower instruction set architecture (ISA), formerly most familiar as PowerPC used in Apple computers till the mid 00s. One to follow (as far as Framework users are concerned) is Power Progress Community’s PPC Notebook. There is also Libre SOC who are taking the ISA in new directions for embedded and mobile use. The comeback alluded to in the thread title seems already underway.
I understand the principle benefit of PowerPC over x86, ARM and even RISC-V is that there is no closed source IP whatsoever. All other things being equal, the efficiency of chips built on OpenPower ought to be better than x86 and nearer to other RISC ISAs like ARM. Arguably the ISA is also more mature and ready for wider adoption than is RISC-V.
As an armchair enthusiast, I shall enjoy watching the various OpenPower projects develop. I was an iBook G4 owner back in the day so have fond memory of owning a PowerPC notebook.
Would others agree that the OpenPower revival seems a good fit with Framework’s mission? I understand PowerPC Notebook have found a case for their first model, a 15.6" case. They previously instigated the design of an open source 3D printable laptop of their own, also 15.6". I suggest that whatever the success of these projects to develop a large notebook, using Framework’s chassis would be the best way to produce a smaller PPC notebook for those who like the idea of fully open architecture but have more modest requirements.
Do others share my enthusiasm for the prospects of OpenPower in comparison to other ISA’s? Framework have publicly stated aspirations to bring ARM and RISC-V to their laptop, has there been any consideration given to supporting OpenPower?
I can understand where you got that opinion. Apple sold the transition from PPC to x86 as a logical decision response to Power PC suppliers not delivering the performance gains that had been expected. Arguably company politics were a bigger factor than the merits of the chips or their underlying design philosophy. I rather suspect Apple didn’t suddenly decide the x86 architecture was intrinsically better than that of PowerPC. Rather the opposite, I suspect, but accepted the less elegant solution on commercial grounds.
Of course the Apple enthusiasts will tell us that the switch to Intel was a farsighted masterstroke, bringing vastly better performance plus Bootcamp for Windows, and driving the exponential growth of Mac sales. Of course this massive growth may well have happened anyway as the Apple “ecosystem” expanded, but we will never know.
I disagree with your assumption the architecture is somehow only suitable for “stationary” (server?) applications. PowerPC underpinned consoles Xbox360, Playstation3 and Wii, and embedded microcontrollers have been built around the ISA. No more so than Arm (which debuted on a desktop computer) being limited to that application.
Anyway, that was then. Today it is Intel who are struggling to get more performance out of their chips. PowerPC on the other hand is open for others to take in new directions. Clearly the Power Progress Community wold disagree that architecture is not suitable for laptop use. You might also want to glance at how previously linked Libre-SOC is taking the ISA in new directions specifically for embedded use.
Hmm, interesting question. My main experience was with RS/6000 POWER servers and workstations, which kinda-sorta influenced me to acquire and mess around with various and sundry PPC computers back in the day (the G3 iMac, a TI Powerbook G4, a couple of older 601-powered Macs with aftermarket G3 and G4 processor cards, and even a couple of Motorola Mac clones) but nothing recently. The state of PPC linux back then was perhaps different than it is now but it wasn’t ready for prime-time for me back then.
If folks have been working to bring the ISA and architecture forward for modern mobile that would be a good thing; most of the development I’ve stumbled across recently seems to be for embedded systems, many of which seem to be rocking the same PPC 750 G3 that were used in the old days.
I don’t manage big-iron AIX servers any more so I don’t really know what the state of POWER is with regards to lithography and clocks, but I can’t imagine IBM hasn’t been working hard to keep them comparable to other server processors as far as performance per watt and overall performance.
I like the idea, but I’d need solid desktop-style support (e.g., not just a framebuffer running a shell) in an OS for my interest to really be piqued. Maybe that’s already out and I’m just unaware. That the PPC Notebook is using an MXM card at least means they’re intending to use something other than good old framebuffer which is a good sign.
I got started in the IT world when there were tons of architectures (DEC Alpha, PA-RISC, SPARC, MIPS, POWER, ia32, etc.), but the cheap price and easy availability of ia32 and eventually amd64 killed off many of those for lower-end applications, and that kind of killed off designs that could evolve into consumer-style desktops or workstations, I think. We’ll have to see! If they can get at least some commodity (for system integrators) drivers that aren’t just binary blobs so folks have good driver support for commonly-desired devices (graphics cards, wireless radios, ethernet controllers, sound cards, USB bus devices, and so forth) maybe it could be an option.
Raptor Computing Systems have a range of POWER9 based hardware which may be of interest. News on Raptor products and OpenPower topics generally can be found on Talospace (“The free computing Frontier”). There seems to be a lot going on in the OpenPOWER scene even if it does not enjoy the same profile as other platforms.
I agree, it is very sad that a lot of these architectures were lost when it seemed like Itanium would be the future. DEC Alpha seems the most tragic of those that got away. I understand that as the most recent RISC design of that era DEC was able to make theirs the best and unlike all the others (including POWER) it was 64 bit from the outset. Windows NT had been ported to the architecture and it seemed to have had a good future before it. I can’t help but wonder how things might have panned out differently if it had remained on the market.
It’s certainly looking that way. At some point in the last two weeks or so. RED Semiconductor’s website came up from being parked. For those not in the know, RED is libre-SoC’s fab company they set up recently. With any luck, they should get a product out by 2023 or so, engineering is still finalizing their SimpleV vector instruction set extensions though so '24 isn’t out of the question.
Hopefully in making a reference board, they consider making one based on the released Framework board sources.